Solo
$19/ month
750 credits
For independent builders who want a reliable finish check before they ship.
- Shared finish-credit pool
- Every standard verifier model
- Local + cloud finish runs
PLANS + VERIFIER CATALOGPay for the
Pay for the
check, not seats.
Shared credits for independent finish checks. Choose the verifier models, see their cost, and spend the same pool across your team.
FINISH ECONOMICSLIVE CATALOG
from0.5CREDITS / VERIFIER RUN
- Models
- 33
- Team lanes
- 3
- Free evaluation
- 100 cr
NO PER-SEAT TAXVERIFIER CHOICE STAYS VISIBLE100 CREDITS TO EVALUATE
01 / BUNDLED CREDITSChoose the size
Choose the size
of the safety net.
Every paid pool is shared. Start small, move up when finish volume rises, or bring your own provider key.
Pro
RECOMMENDED$49/ month
3,000 credits
The default for teams that finish-check every meaningful agent change.
- Larger shared credit pool
- Repair packages on fail
- Frontier verifier escalation
Team
$149/ month
10,000 credits
Higher finish volume, shared controls, and frontier escalation when needed.
- Highest bundled capacity
- GPT-5.6 Sol + Claude Opus 5
- Priority finish support
04BYOK
DIRECT MODEL BILLING
$8 / month
Critique runs the finish harness. Your LLM Gateway, CrofAI, or OpenRouter account pays for verifier tokens directly.
- E2B sandboxes + finish orchestration
- Retrieval, logs + secure key storage
Annual billing discounts are available at checkout. Verified students and open-source maintainers can apply for the $5/month lane.
02 / HOW CREDITS MOVENo mystery
No mystery
routing bill.
You choose the verifier and optional specialist models. Critique shows the floor before the run; change depth scales the final cost.
Open cost calculator →01Verifiercredits
+02Specialistscredits
×03Change depth0.75–2×
=04Finish costvisible
03 / STARTING ROUTESStart with a stack.
Start with a stack.
Change any layer.
One model can handle every role, but mixed stacks let you spend frontier credits only where they improve the finish decision.
PROFILEVERIFIERSPECIALISTREPAIREST. COST
01Best default
GLM-5.2Qwen3.7 PlusMiMo v2.5 Pro3–8 cr02Cheap volume
DeepSeek V4 FlashLing-2.6 FlashMiMo v2.51–3 cr03Frontier escalation
Claude Sonnet 5GPT-5.6 TerraClaude Opus 540–240+ cr04 / MODEL LEDGEREvery model.
Every model.
One clear floor.
All standard models can work as verifier, specialist, or repair path. Team lanes unlock the heaviest escalation models.
MODELPROVIDERCONTEXTSPEEDACCESSFLOOR
01DeepSeek V4 Flash
DeepSeek1MFastSolo + Pro0.5 cr02Gemma-4-31B
Google262KFastSolo + Pro0.5 cr03Hy3 Preview
Tencent262KBalancedSolo + Pro0.5 cr04Ling-2.6-Flash
InclusionAI262KInstantSolo + Pro0.5 cr05MiMo v2.5
Xiaomi1MInstantSolo + Pro0.5 cr06DeepSeek V4 Pro
DeepSeek1MBalancedSolo + Pro1 cr07Hy3
Tencent262KBalancedSolo + Pro1 cr08MiMo v2.5 ProDEFAULT
Xiaomi1MDeepSolo + Pro1 cr09Qwen3.7 Flash
Alibaba131KBalancedSolo + Pro1 cr10StepFun-3.7 FlashDEFAULT
StepFun262KInstantSolo + Pro1 cr11Trinity-Large-Thinking
Arcee AI262KBalancedSolo + Pro1 cr12MiniMax-M2.7
MiniMax197KBalancedSolo + Pro1.5 cr13Qwen3.7 Plus
Alibaba1MBalancedSolo + Pro1.5 cr14Ring-2.6-1T
InclusionAI262KBalancedSolo + Pro1.5 cr15Gemini 3.1 Flash Lite
Google1MInstantSolo + Pro2 cr16KAT Coder Pro v2.5
KwaiPilot256KFastSolo + Pro2 cr17GLM-5.2DEFAULT
Z.AI1MBalancedSolo + Pro3 cr18Grok Build 0.1
xAI256KBalancedSolo + Pro3 cr19MiniMax-M3
MiniMax1MBalancedSolo + Pro3 cr20Nemotron 3 Ultra
NVIDIA1MDeepSolo + Pro3 cr21Gemini 3.6 Flash
Google1MFastSolo + Pro4 cr22Kimi K2.6
MoonshotAI262KBalancedSolo + Pro4 cr23Kimi K2.7 Code
MoonshotAI262KDeepSolo + Pro4.5 cr24Grok 4.5
xAI1MBalancedSolo + Pro6 cr25Qwen3.7-Max
Alibaba262KBalancedSolo + Pro6 cr26Gemini 3.5 Flash
Google1MBalancedSolo + Pro10 cr27Claude Sonnet 5DEFAULT
Anthropic1MBalancedSolo + Pro15 cr28Gemini 3.1 Pro
Google1MDeepSolo + Pro16 cr29GPT-5.6 Terra
OpenAI1MBalancedSolo + Pro20 cr30Kimi K3
MoonshotAI131KBalancedSolo + Pro20 cr31Claude Opus 5
Anthropic1MDeepTeam30 cr32GPT-5.6 Sol
OpenAI1MDeepTeam40 cr33GPT-5.6 Sol Pro
OpenAI1MBalancedTeam55 cr05 / THE FINE PRINTStraight
Straight
answers.
01What counts as one credit?+
A credit is Critique’s normalized unit for a finish check. Verifier and optional specialist model costs are added together, then scaled by the size and depth of the change. A typical finish pass is commonly 3–50+ credits.
02Do unused credits roll over?+
No. Credits reset at each monthly billing boundary. The model ledger makes costs visible so you can choose the pool that matches your actual finish volume.
03Can one model do every job?+
Yes. Every finish-capable model can work as the primary verifier, a specialist pass, or the repair path within its plan gate. The suggested stacks are starting points, not locked bundles.
04How does BYOK change billing?+
The $8 monthly harness pays for Critique orchestration. LLM Gateway, CrofAI, or OpenRouter bills model tokens to your provider account, while Critique continues to run sandboxes, checks, retrieval, and usage ledgers.